Fire Detection

Author: Pedro Vinícius A. B. Venâncio¹

¹ Graduate Program in Electrical Engineering (PPGEE/UFMG)

---

About

This repository contains the models and source codes of hybrid systems for fire detection implemented during my master's degree, as well as some baseline models for comparison purposes. The proposed hybrid systems are composed of two sequential stages: (i) spatial detection, which consists of identifying and locating fire and smoke events on the scene based on spatial patterns, and (ii) temporal analysis of the events detected in the previous stage, in order to make a final decision on whether a fire is actually taking place. The baseline models are simple convolutional neural networks for fire classification proposed in the literature.

How to run fire and smoke detection on a video

Tutorial

1. After cloning the repository, copy the videos you want to run the algorithms to the examples/ folder.

2. Build the fire-detection image from the available Dockerfile.

docker build -t fire-detection .

3. Create and run a new container from the fire-detection image.

docker run -it --rm fire-detection /bin/bash

4. Choose which model you want to run and follow the steps in its respective subsection.

Detection using a hybrid system

The first stage of the hybrid system is a YOLOv5 network (small or large) and the second stage can be a area variation technique (AVT) or a temporal persistence technique (TPT). We recommend AVT for outdoor scenes and TPT for indoor scenes.

After running the system, the videos with the detections are saved in runs/detect/exp/.

YOLOv5+AVT

If you want to use the hybrid system YOLOv5+AVT, run the following command inside the container:

python detect.py --source <video_file> --weights ./weights/<weights_file> --temporal tracker

where <video_file> is the video in which you will detect fire and <weights_file> is the file with the network weights (can be yolov5s.pt or yolov5l.pt). You can change the parameters of the area variation technique by specifying the additional flags --area-thresh and window-size.

YOLOv5+TPT

If you want to use the hybrid system YOLOv5+TPT, run the following command inside the container:

python detect.py --source <video_file> --weights ./weights/<weights_file> --temporal persistence

where <video_file> is the video in which you will detect fire and <weights_file> is the file with the network weights (can be yolov5s.pt or yolov5l.pt). You can change the parameters of the persistence temporal technique by specifying the additional flags --persistence-thresh and window-size.

Detection using YOLOv5

YOLOv5

If you want to use only the YOLOv5 network, run the following command inside the container:

python detect.py --source <video_file> --imgsz 640 --weights ./weights/<weights_file>

where <video_file> is the video in which you will detect fire and <weights_file> is the file with the network weights (can be yolov5s.pt or yolov5l.pt). You can change the parameters of the YOLOv5 network by specifying the additional flags --img-size, --conf-thres and --iou-thres.

Detection using baseline models

If you want to use a baseline model, run the following command inside the container:

python baseline.py --video <video_file> --model <model_name>

where <video_file> is the video in which you will detect fire and <model_name> is the name of the model to be used (can be 'firenet' or 'mobilenet').

Models

Download the model weights from the root of this repository by running the ./scripts/download_models.sh script or manually using the links below.

• Baseline models.
• Proposed models.

Citation

Please cite the following paper if you use our proposed hybrid systems for fire and smoke detection:

• Pedro Vinícius Almeida Borges de Venâncio, Roger Júnio Campos, Tamires Martins Rezende, Adriano Chaves Lisboa, Adriano Vilela Barbosa: A hybrid method for fire detection based on spatial and temporal patterns. In: Neural Computing and Applications, 2023.

If you use our YOLOv4 models for fire and smoke detection, please cite the following paper:

• Pedro Vinícius Almeida Borges de Venâncio, Adriano Chaves Lisboa, Adriano Vilela Barbosa: An automatic fire detection system based on deep convolutional neural networks for low-power, resource-constrained devices. In: Neural Computing and Applications, 2022.

References

• Algorithms: YOLOv4 with Darknet, YOLOv5 with PyTorch, and Object tracking with OpenCV.
• Baseline models: Inferno fire detection using CNNs, and FireNet lightweight network for fire detection.
• Datasets: D-Fire dataset, FireNet dataset, and Foggia's dataset.